Overview

This plugin adds (DeepSeek V4.1 Flash) alongside , , and through the official Chat Completions API.

Select DeepSeek V4.1 Flash for text and image input with reasoning, tool calling, and JSON Output.
It uses the same API key and base URL, with a 1M-token context and up to 393,216 output tokens, including reasoning.
The new V4.1 Flash entry defaults to thinking mode with effort and a 65,536-token output limit.
Use the Thinking Mode switch to disable reasoning, or choose , , or effort when it is enabled.
For the new V4.1 Flash entry, applies only in thinking mode, from 0.95 to 1.0; applies only in non-thinking mode.
See the Chat Completions reference and Thinking Mode guide for parameter behavior.

Images can be supplied as URLs or inline Base64 in user and tool messages.
Supported image formats are JPEG, PNG, GIF, and WebP; see the official vision guide for size and request limits.
Local token estimates exclude images; the API's returned token usage includes them.

Existing Dify apps can keep their selected model after upgrading; no model migration is required.
The , , and entries retain their existing IDs, labels, capabilities, parameter defaults, accepted ranges, and sampling behavior.
DeepSeek still accepts both legacy Flash API names and routes them to V4.1 Flash; this plugin sends the selected model ID unchanged.
Credential validation continues to use the existing API name.
Select DeepSeek V4.1 Flash when configuring the new model explicitly.
remains available; from September 14, 2026 at 04:00 UTC, DeepSeek will route its requests to V4.1 Flash until V4.1 Pro launches.
See the official release notes for the migration schedule.

DeepSeek pricing varies by cache status and time window, so refer to the official pricing page for current rates.

Configure

Get an API key from DeepSeek, then configure it under Settings → Model Provider.